从文本特征项所处的位置角度提出了特征项基于位置的降维方法;同时结合特征的类别分布进行了二次特征降维。这种基于位置和类别相结合的特征降维方法在最大程度减少信息损失的条件下,实现了特征维数的有效压缩。实验表明,该方法有较高的文本分类效率。
From the position of the terms, this paper put forward a method to reduce the dimensionality. Meanwhile, combined with the sorts distributing, it once more reduced the feature dimension. Therefore, in precondition of the information loss least, connecting with the two aspects, used this method to complete the text feature decrease smartly. The test shows that this method has better precision in the text categorization.